检索结果-内蒙古大学图书馆

IEEE International conference on Big Data (Big Data)

作者： Zhang, Lingxin Qi, Qi Wang, Jingyu Sun, Haifeng Liao, Jianxin Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9781728108582

The rapid development of artificial intelligence in multiple scenarios, including machine learning, image recognition, and autonomous driving, has led to an explosion of computation jobs. These jobs are often divided into parallel child tasks and executed in distributed clusters with limited computing resources, making parallel task scheduling one of the most important research topics nowadays. Most studies about parallel task scheduling focused on formulating special scenarios and service requirements as optimization problems. However, the complicated and dynamic parallel computing environments are hard to model, predict and control, making those previous methods unscalable and unable to reflect the real scenarios. In this paper, a Multi-task Deep reinforcement learning approach for scalable parallel Task Scheduling (MDTS) is firstly devised. Generally, Deep Reinforcement Learning (DRL) is a modelfree optimization algorithm for long-term control by leveraging experience, but it suffers the curse of dimensionality for decision when coping with complex parallel computing environments and jobs with diverse properties. We extend the action selection in DRL to a multi-task decision, where the output branches of multitask learning are fine-matched to parallel scheduling tasks. Child tasks of a job are accordingly assigned to distributed nodes without any human knowledge while the resource competition among parallel tasks is leveraged through shared neural network layers. Extensive experiments show that the MDTS significantly reduces the job execution lime compared with least-connection scheduling and particle swarm optimization algorithm by 153% and 39.8% respectively. Moreover, MDTS outperforms the raw DRL algorithm on job execution time, load imbalance value, and total cost by 42.8%, 47.5%, and 59.0%.

关键词： parallel task scheduling deep reinforcement learning multi-task learning

来源：评论

学校读者我要写书评

暂无评论

The Internet of Things and fast data streams: prospects for geospatial data science in emerging information ecosystems

引用

CARTOGRAPHY AND GEOGRAPHIC INFORMATION SCIENCE 2019年第1期46卷 39-56页

作者： Armstrong, Marc P. Wang, Shaowen Zhang, Zhe Univ Iowa Dept Geog & Sustainabil Sci Iowa City IA 52242 USA Univ Illinois Dept Geog & Geog Informat Sci Urbana IL 61801 USA

This paper surveys the rapid development of the Internet of Things, the massive data streams that are only now beginning to be generated from it, and the resulting opportunities and challenges that these data streams bring to geographic information analysis. These challenges arise because streaming data volumes cannot bt subjected to analysis using the standard repertoire of methods that have been designed to analyze static geospatial datasets. New approaches are needed, not to supplant, but to supplement, these existing tools. A focus is placed on the concept of data velocity (fast data) and its effects on sampling and inference. Innovative data ingestion strategies based on principles related to reservoir sampling and sketching are described. Dynamic temporal data flows present significant challenges to load balancing in distributed (e.g. cloud) parallel environments, even at exascale levels of performance. Further advances in the exploitation of data locality based on geographical concepts, as well as advanced processing methods based on edge and approximate computing, require further elucidation. Concepts are illustrated using a database compiled from a distributed sensor network of mobile radioactivity detectors.

关键词： Internet of Things sensors data streams fast data parallel computing

来源：评论

学校读者我要写书评

暂无评论

Fuzzy Classification on the Base of Convolutional Neural Networks 2nd

Fuzzy Classification on the Base of Convolutional Neural Net...

引用

2nd International conference of Artificial Intelligence, Medical Engineering, Education (AIMEE)

作者： Puchkov, A. Dli, M. Kireyenkova, M. Natl Res Univ Moscow Power Engn Inst Branch Smole Energetichesky Proyezd 1 G Smolensk 2014013 Russia

ISBN: (纸本)9783030120825;9783030120818

The paper deals with the algorithm of object classification based on the method of fuzzy logic and the application of artificial convolutional neural networks. Every object can be characterized by a set of data presented in the numerical form and in the form of images (photographs in different parts of the light spectrum). In this case, one object can be matched with a few images associated with it;they can be received by different methods and from different sources. In the algorithm, this generalized totality of images is recognized by convolutional neural networks. A separate neural network is formed for every channel of data receiving. Then, the network outputs are combined for processing in the system of classification on the basis of fuzzy logic output. The normalized outputs of convolutional neural networks are used as values of a membership function to terms of outputs variables when a fuzzy classifier works. For the first adjustment of the convolution neural network hyperparameters, the gradient method is applied. The algorithm is realized in Python language with the use of Keras deep learning library and Tensor Flow library of parallel computation with CUDA technology from NVIDIA company. This paper presents the results of practical application of the developed neuro-fuzzy classifier to forecast the problem of working time losses.

关键词： Fuzzy logic Classification image recognition Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Alzheimer's disease detection and classification using machine learning techniques

Alzheimer's disease detection and classification using machi...

引用

4th Smart Cities Symposium (SCS 2021)

作者： M. Bachute N. Vyas S. Khanna K. Modi C. Katpatal

In research work,Alzheimer's disease detection and classification algorithm are presented. We studied the hippocampus region from MRI scans of around 127 elderly patients with AD and 114 scans of people with No cognitive impairment. The areas of brain we examined were Cornu Ammonis (CA) which is subdivided in four successive parts (CA4 to CA1) and the Subiculum at the base end, Stratum radiatum (SR), Stratum lacunosum (SL) and the Stratum molecuare (SM). The final image we generate to extract the visual texture features is a step which occurs much later [1-2]. The first step is to acquire the dataset from ADNI USC depository. After that we submit the files on VolBrain to generate an MRI from our choice of segmentation protocol which gives us our hippocampus. After uploading the neuroimaging file and the Volume of interest upon it in the BrainSuite software we get our grey area of the hippocampus which then can be utilized for algorithms [2-3]. So, to prove the efficacy of texture-built features such as gray level co-occurrence matrix (GLCM) which creates a diagonally distributed co-occurrence matrix. Feature extraction is necessary for several machine learning algorithms to sustain a greater competence. Texture analysis produces an exceptional form of characteristics which assist in delivering us with precise classification and segmentation. We considered a total of three characteristics to disintegrate the data which was Age, Left and Right Hippocampus volumes which led us to various discoveries about how to go about the classification of data. Now we deployed a number of models to our data after it was preprocessed but to select the best fitting algorithm, we had to use F-test which compared these statistical patterns to fit a data set in order to distinguish the best fit. After applying GLCM, numerous other methods of feature extraction were used like Correlation which calculates the odds of a co-occurring couple of associated pixels. Decision trees was another c

关键词： biomedical MRI brain cognition decision trees diseases feature extraction image classification image recognition image representation image segmentation image texture medical image processing neurophysiology statistical analysis unsupervised learning nearest neighbour methods

来源：评论

学校读者我要写书评

暂无评论

A Comparison Study of GRAPPA and Generalized Series methods for parallel MRI at high acceleration factor

A Comparison Study of GRAPPA and Generalized Series Methods ...

引用

Annual Summit and conference of the Asia-Pacific-Signal-and-Information-processing-Association (APSIPA ASC)

作者： Nguyen, Khanh T. K. Nguyen, Hien M. Vietnamese German Univ Binh Duong Vietnam

ISBN: (纸本)9781728132488

A comparison study of GRAPPA and Generalized Series for parallel MRI images is performed and analyzed at high acceleration factor. Through extensive simulation experiments on various data sets, the conventional GRAPPA method proved its efficiency at low acceleration factor when acquiring few auto-calibrating lines. However, at high acceleration factor 8-12 with very sparse k-space sampling, the GS method proved to be more superior by significantly reducing noise level and achieving much higher accuracy. The key to success of GS method is sufficient amount of low frequency info contained in reference image. This can be achieved by collecting a sufficient the number of ACS lines. In this study, 15%-20% of total k-space lines was sufficient.

关键词： pMRI undersample GRAPPA GS reconstruction method high acceleration factor

来源：评论

学校读者我要写书评

暂无评论

Performance evaluation of the optical navigation electronics of HAYABUSA2

Performance evaluation of the optical navigation electronics...

引用

SPIE conference on Future Sensing Technologies

作者： Hihara, Hiroki Sano, Junpei Masuda, Tetsuya Ootake, Hisashi Okada, Tatsuaki Ogawa, Naoko Tsuda, Yuichi NEC Space Technol Ltd 1-10 Nisshin Cho Fuchu Tokyo 1838551 Japan NEC Corp Ltd 1-10 Nisshin Cho Fuchu Tokyo 1838501 Japan Japan Aerosp Explorat Agcy JAXA 3-1-1 Yoshino Dai Sagamihara Kanagawa 2525210 Japan

ISBN: (数字)9781510631120

ISBN: (纸本)9781510631120

Digital Electronics and Optical Navigation Camera (DE-ONC) is an edge computing node of the asteroid probe HAYABUSA2. DE-ONC was developed to provide real-time image recognition performance for optical navigation. Lightweight, low power consumption and miniaturization are realized to overcome resource restrictions. It also satisfies high reliability and safety requirements of HAYABUSA2 missions. There are static and dynamic requirements for reliability and safety. The former increases reliability by adding redundancy combining the concept of functional distribution and time-division redundancy to meet resource constraints. Functional distribution mode, standby redundancy mode and hot redundancy mode were realized with the same device configuration. The real-time performance of optical navigation exploiting image recognition functions of the unit was demonstrated through the interplanetary cruising phase, as well as touch down to and taking off from the asteroid Ryugu. DE-ONC is always required to operate in the critical operation phase. In addition to that, it must always satisfy latency requirements to complete processing within a predetermined duration and to guarantee hard real-time performance. In order to satisfy these requirements, the image processing unit of DE-ONC adopts a unified language processing system and a distributed memory model with reference to a parallel inference machine, which is a so-called the second generation artificial intelligence technology. Its image processing module integrates a radiation hardened micro-controller unit (MCU) and field programmable gate arrays (FPGAs) with the language processing system and the distributed object model. We report the evaluation result of reliability and safety with real-time performance of the unit's architecture.

关键词： SpaceWire image recognition real-time optical navigation networking spacecraft electronics internet of things (IoT) artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Color distortion removal for heart rate monitoring in fitness scenario 20

Color distortion removal for heart rate monitoring in fitnes...

引用

20th International conference on parallel and distributed Computing, Applications and Technologies, PDCAT 2019

作者： Tran, Quoc-Viet Su, Shun-Feng Tran, Minh-Quang Department of Electrical Engineering National Taiwan University of Science and Technology Taipei106 Taiwan Department of Mechanical Engineering National Taiwan University of Science and Technology Taipei106 Taiwan

ISBN: (纸本)9781728126166

Heart rate estimation from fitness plays an important role in the evaluation of fitness exercises. Conventional approaches use the photoplethysmography (PPG) sensor to consider the change of light absorption on the wrist skin for heart rate estimation. However, users are required to buy smartwatches for using this function. Various approaches based on video analysis are recently implemented for surveillance purpose. However, it is unstable for motion scenario such as fitness exercises due to the color distortion induced by movement. POS and CHROM are introduced to address this issue. Since the fixed projection planes from POS and CHROM are given in several sources of light, it is not widely applied for surveillance applications. Therefore, a novel projection plane that is adaptively changed with the lighting environment is proposed to estimate the heart rate from fitness videos in ambient light. Moreover, image and digital signal processing techniques are also applied to extract the clean pulse signal from a novel projection plane. From the experiments conducted, the proposed approach outperformed the existing approaches to be the best model for heart rate estimation from fitness videos with the accuracy up to 91.08%. © 2019 IEEE.

关键词： Digital signal processing

来源：评论

学校读者我要写书评

暂无评论

DSPG: Decentralized Simultaneous Perturbations Gradient Descent Scheme

DSPG: Decentralized Simultaneous Perturbations Gradient Desc...

引用

Euromicro conference on parallel, distributed and Network-Based processing

作者： Arunselvan Ramaswamy Dept. of Electrical and Computer Science Engineering Paderborn University Paderborn Germany

ISBN: (数字)9781728165820

ISBN: (纸本)9781728165837

distributed descent-based methods are an essential toolset to solving optimization problems in multi-agent system scenarios. Here the agents seek to optimize a global objective function through mutual cooperation. Oftentimes, cooperation is achieved over a wireless communication network that is prone to delays and errors. There are many scenarios wherein the objective function is either non-differentiable or merely observable. In this paper, we present a cross-entropy based distributed stochastic approximation algorithm (SA) that finds a minimum of the objective, using only samples. We call this algorithm Decentralized Simultaneous Perturbation Stochastic Gradient, with Constant Sensitivity Parameters (DSPG). This algorithm is a two fold improvement over the classic Simultaneous Perturbation Stochastic Approximations (SPSA) algorithm. Specifically, DSPG allows for (i) the use of old information from other agents and (ii) easy implementation through the use simple hyper-parameter choices. We analyze the biases and variances that arise due to these two allowances. We show that the biases due to communication delays can be countered by a careful choice of algorithm hyper-parameters. The variance of the gradient estimator and its effect on the rate of convergence is studied. We present numerical results supporting our theory. Finally, we discuss an application to the stochastic consensus problem.

关键词： Sensitivity Approximation algorithms Delays Linear programming Convergence Gradient methods

来源：评论

学校读者我要写书评

暂无评论

Digital Signal processing Based on Spline Functions

Digital Signal Processing Based on Spline Functions

引用

2019 International conference on Information Science and Communications Technologies, ICISCT 2019

作者： Mirzaev, Avaz Khalilov, Sirojiddin Information Technologies Dept Tashkent Uzbekistan

ISBN: (纸本)9781728125640

The development of the modern world's information and communication technologies and their implementation in various industries are important, are essential to the creation of parallel algorithms for solving problems of the recovery and digital signal processing, on the basis of the treatment processes a multi-core architecture and the search for optimal solutions. The current stages of development of the structure of machines, complexes and systems installed at facilities that are part of the structure of devices specialized for conducting scientific research or located on rolling stock are characterized by on-line analysis of complex processes and fields and increased demands on the speed of processing large amounts of data in real time. Particular attention is paid to the development of methods for solving problems of image and signal processing by using modern methods, algorithms and structural tools, architecture of computing and software. In that work the result of the study of methods of approximation of functions and functional dependencies obtained in the course of experiments for the selected class of signals improved method of spline functions on the basis of analytical analysis, parallel algorithms for digital signal processing have been developed, efficient algorithms have been developed for digital processing of vibrio signals in multi-core processors based on cubic and bi cubic basis splines. © 2019 IEEE.

关键词： Digital signal processing

来源：评论

学校读者我要写书评

暂无评论

Fast, Sub-pixel Accurate Digital image Correlation Algorithm Powered by Heterogeneous (CPU-GPU) Framework

Fast, Sub-pixel Accurate Digital Image Correlation Algorithm...

引用

Society-for Experimental-Mechanics (SEM) Annual conference and Exposition on Experimental and Applied Mechanics

作者： Thiagu, Mullai Subramanian, Sankara J. Nasre, Rupesh Indian Inst Technol Madras Dept Engn Design Chennai Tamil Nadu India

ISBN: (纸本)9783319974811;9783319974804

Digital image Correlation (DIC) is a popular non-contact image-based full-field deformation measurement tool widely used in mechanics. In spite of its significant advantages, it is still primarily used as a post-processing tool due to its computational cost. In recent years, parallel computing platforms such as multi-core processors and Graphics processing Units (GPUs) have been used to improve the speed of the DIC algorithm, with GPUs being well-suited for implementing data-parallel operations. Previous works have performed GPU-based DIC wherein each sub-image (i.e. a collection of a few pixels in the local neighborhood of a point of interest) is allocated to a single thread on the GPU, thus achieving parallelism across sub-images. However, this is not the only type of parallelism that is possible: one can also achieve parallelism within a sub-image as well as across whole images. The aim of this work is to efficiently implement 2D-DIC such that parallelism within a sub-image as well as across sub-images leads to considerable reduction in computation time. We use a heterogeneous framework consisting of an Intel Xeon octa-core CPU and an Nvidia Tesla K20C GPU card in this work. The CPU is used to handle image pre-processing, whereas the GPU is used to process four compute-intensive tasks: affine shape function computation, B-Spline interpolation, residual vector calculation and deformation vector update. parallelization within and across sub-images is achieved in this work by efficient thread handling and use of pre-compiled BLAS libraries. In order to estimate the speedup provided by the GPU, the same four tasks were also evaluated on the octa-core CPU;a speedup of approximately 7 to 5 times was observed for a single sub-image whose size varies from 21x21 to 61x61 respectively. However, it is expected that for a larger number of sub-images, the GPU speedup will be higher and this is indeed the case: when the affine shape function computation and B-Spline interpolat

关键词： Full-field displacement Sub-image parallel computing Heterogeneous framework Compute Unified Device Architecture (CUDA) Thread Kernel

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：